Load dataset

Let’s read in our dataset.

library(tidyverse)

d <- read_csv("./Data/data.csv")

Review Practice Problems

  1. Create a box plot with the x-axis as year.

First, let’s try the obvious plot, using x = year, y = MOR, and adding geom_boxplot().

However, running this we get a weird result: why?

ggplot(d, aes(x = year, y = MOR)) +
  geom_boxplot()

This is a formatting issue because our dataset has year as an integer, not a category.

What’s an easy way to remediate this problem? Transforming year to a factor.

ggplot(d, aes(x = as.factor(year), y = MOR)) +
  geom_boxplot()

  1. Next, use a facet layer to trellis the box plots by region so that each region has its own temporal box plots. Then create a new plot with only Sub-Saharan Africa and South Asia.

For the next part, we can facet to create a plot for each region.

ggplot(d, aes(x = as.factor(year), y = MOR, color = region)) +
  geom_boxplot() + facet_wrap(~region)

The chart is still too busy. Perhaps we want to look at only Sub-Saharan Africa and South Asia.

reg <- c("Sub-Saharan Africa", "South Asia")

ggplot(d[d$region %in% reg,], aes(x = as.factor(year), y = MOR, color = region)) +
  geom_boxplot()

  1. Clean up the plot with Sub-Saharan Africa & South Asia by adding x, y, and title labels; moving the legend to the bottom; and color each region by a unique color.
ggplot(d[d$region %in% reg,], aes(x = as.factor(year), y = MOR, fill = region)) +
  geom_boxplot() + 
  theme(plot.title = element_text(hjust = 0.5), legend.position = "bottom") +
  labs(x = "Year", y = "Mortality Rate Distribution", title = "Sub-Saharan Africa & South Asia: Mortality Rates") +
  scale_color_discrete(guide = guide_legend(title = "Region"))

themes and ggthemes

Themes can also be helpful to modify the overall style of a plot.

There are helpful add-ins within the ggthemes package, which can provide a broader range of templates.

For example: Tufte Box Plots (“Minimum Ink”)

First, let’s rerun the data manipulations we did yesterday…

pop2014 <- d[d$year==2014,]
pop2014$region <- factor(pop2014$region, 
                         levels = c("Sub-Saharan Africa",
                                    "South Asia",
                                    "East Asia & Pacific",
                                    "Latin America & Caribbean",
                                    "Middle East & North Africa",
                                    "Europe & Central Asia",
                                    "North America"
                                    ))

levels(pop2014$region) <- gsub("&", "\n &", levels(pop2014$region))

Now, let’s create the plot (after calling ggthemes)…

#install.packages("ggthemes")
library(ggthemes)

g <- ggplot(pop2014, aes(region, MOR)) + labs(x = "Region", y = "Mortality Rate") 
g

change the theme…

g + theme_tufte(ticks=FALSE)

then add the geom…

g + theme_tufte(ticks=FALSE) +
  geom_tufteboxplot()

or modify it…

g + theme_tufte(ticks=FALSE) +
  geom_tufteboxplot(median.type = "line", whisker.type = 'line')

g + theme_tufte(ticks=TRUE) +
  geom_tufteboxplot(median.type = "line", whisker.type = 'line', hoffset = 0, width = 2)

There are many other themes like…

#recall from yesterday's code
years <- c(2001, 2005, 2009, 2013)
d1 <- d[d$year %in% years,]

p <- ggplot(d1, aes(x = HEX, y = MOR, color = region)) +
  geom_point() +
  facet_wrap(~year) + 
  scale_x_log10() +
  scale_y_log10() +
  labs(x = "Health Expenditure as a Percent of GDP (Log)", y = "Mortality Rate (Log)") +
  geom_smooth(method=lm, se=FALSE)  # Don't add shaded confidence region

p + theme_solarized() +
  scale_colour_solarized("blue") +
  scale_color_discrete(guide = guide_legend(title = "Region"))

p + theme_igray()

p + theme_few() + 
  scale_colour_few() + 
  scale_color_discrete(guide = guide_legend(title = "Region"))

Here’s a full list: https://github.com/jrnold/ggthemes

ggplot2 for plotly

Plotly is a package of interactive visualizations.

In fact, they have a ggplot2 package called plotly that is a wrapper function ggplotly on ggplot2 but for interactive visualizations.

There are multiple ways to run interactive ggplot, but the easiest is to use the same logic for ggplot but then pass the object to the function ggplotly().

#install.packages("plotly")
library(plotly)

ggplotly(p)
#rerun data manipulations from yesterday
d$r.MOR <- d$MOR * (d$POP) / 1000
d2 <- d %>% 
  na.omit() %>% # note the na.omit is very important!
  group_by(region, year) %>% 
  summarise(t.MOR=sum(r.MOR),
            t.POP=sum(POP)) %>%
  mutate(MOR.rate = t.MOR / (t.POP / 1000))

p <- ggplot(d2, aes(year, MOR.rate, color = region)) + 
  geom_line() +
  geom_point() + 
  labs(x = "Year", y = "Mortality Rate") +
  scale_color_discrete(guide = guide_legend(title = "Region"))

ggplotly(p)

Web-interactive Visualizations

What’s the range of possible ggplot2 options?

There’s a great open source Shiny app to explore ggplot2: https://www.r-bloggers.com/shiny-app-to-explore-ggplot2/

This visualization utilizes Shiny, which is a R visualization toolkit to build interactive visualizations without knowing JavaScript, HTML, and CSS.

I’ve previously created a R Visualization Tutorial that includes several examples using Twitter data.